Supplementary material for “Submodularity beyond submodular energies: coupling edges in graph cuts”
نویسندگان
چکیده
To investigate the effect of coupling edges, we compare cooperative cut (CoopCut) to the standard graph cut (GraphCut), and, for shrinking bias, also to curvature regularization. To ensure equivalent conditions, all methods used the same weights on the terminal edges (i.e., the same unary potentials), the same 8-neighbor graph structure, and the same inter-pixel edge weights. The unary potentials stem from color histograms [2], or from Gaussian mixture models (GMMs) with 5 components [12, 14]. The weight of an inter-pixel edge e = (vi, vj) ∈ En is w(e) = 2.5 + 47.5 exp(−0.5‖zi − zj‖/σ). (1)
منابع مشابه
Combinatorial problems with submodular coupling in machine learning and computer vision
Numerous problems in machine learning and computer vision are discrete. As a complicating factor, they often involve large data sets and higher-order interactions between elements in the data. For example, segmenting an image into foreground and background requires assigning a label to each pixel in the image. As object and background commonly have significant wide-range coherency, the most pro...
متن کاملWhen can lp-norm objective functions be minimized via graph cuts?
Techniques based on minimal graph cuts have become a standard tool for solving combinatorial optimization problems arising in image processing and computer vision applications. These techniques can be used to minimize objective functions written as the sum of a set of unary and pairwise terms, provided that the objective function is submodular. This can be interpreted as minimizing the $l_1$-no...
متن کاملMulti-label cooperative cuts
Recently, a family of global, non-submodular energy functions has been proposed that is expressed as coupling edges in a graph cut. This formulation provides a rich modelling framework and also leads to efficient approximate inference algorithms. So far, the results addressed binary random variables. Here, we extend these results to the multi-label case, and combine edge coupling with move-maki...
متن کاملSupplementary Material Accompanying ‘Geometry Driven Semantic Labeling of Indoor Scenes’
In this appendix, we will show how the higher order energy potentials can be minimized using graph cuts. Since, graph cuts can efficiently minimize submodular functions, we will transform our higher order energy function (Eq. 9) to a submodular second order energy function. For the case of both αβ-swap and αexpansion move making algorithms, we will explain this transformation and the process of...
متن کاملScaling Submodular Maximization via Pruned Submodularity Graphs
We propose a new randomized pruning method (called “submodular sparsification (SS)”) to reduce the cost of submodular maximization. The pruning is applied via a “submodularity graph” over the n ground elements, where each directed edge is associated with a pairwise dependency defined by the submodular function. In each step, SS prunes a 1 1/ p c (for c > 1) fraction of the nodes using weights o...
متن کامل